Update vid2persona/pipeline/llm.py

vid2persona/pipeline/llm.py

@@ -67,9 +67,11 @@ async def chat(
         "repetition_penalty": repetition_penalty
     }
 
-    if hf_token is None:
+    try:
         for response in local_openllm.send_message(messages, model_id, max_input_token_length, parameters):
             yield response
-    else:
+    except ThreadPoolBuildError as e:
+        gr.Warning("Something went wrong. Switching to TGI remotely hosted model")
+    finally:
         async for response in tgi_openllm.send_messages(messages, model_id, hf_token, parameters):
             yield response